When Honor launched the 400 series last year, it introduced a new AI video generation feature that could turn a static photo into a five-second video clip. It was rather well received, with Honor revealing that users created over 13.4 million seconds of AI videos using the feature. However, it had some obvious limitations.
Honor’s first go at the AI Image to Video feature lacked an option to add specific prompts. So, what the AI decided to do with your image was what you got. The Honor 600 series changes that with AI Image to Video 2.0, which is powered by what Honor calls the industry’s first unified multi-modal video generation model, a system that brings video generation, editing, and comprehension together into a single workflow.
With the new feature, you can now input up to three images alongside your own text prompts to control the outcome. The feature lives in the Gallery app under the Create tab, or you can access it quickly by pressing the dedicated AI button on the Honor 600 Pro.
AI Image to Video 2.0 comes with 19 premade templates, each with its own predefined prompt and style. Cinematic presets cover camera movements like “bullet time” and the “Hitchcock zoom.” There are also motion-based effects, animation templates, and more emotionally driven ones for compositing people into shared moments. If none of them fit, you can skip the templates entirely and write your own prompt to get a customized clip.
How it performs in practice
AI Image to Video 2.0 takes around five to seven minutes to generate clips, so don’t expect instant gratification. But the whole process can run in the background, so your phone stays usable throughout, which is a fair trade-off. I tested a few of the templates and even tried custom prompts, and here’s how it went.
The Drone Pullback template delivered impressive results when I fed it a photo of me standing on the beach. It generated a convincing simulated drone pullback, complete with contextually appropriate audio, and without anything in the original frame being altered.
Magic Wardrobe held up, too. I fed it an image of a mannequin in a striking outfit, and the AI not only swapped the clothing a couple of times, but also animated the mannequin, while leaving the background and shadows intact.
Animation Magic was one of my favorites of the bunch. When applied to a painted landscape with mountains and birds, it animated the existing birds, added new ones in flight, introduced a slow zoom that surfaced additional peaks in the background, and placed a sun in the scene.
That said, some templates worked better than others. The Pet Roleplay template, for instance, is fun in theory but leans hard into a stylised, almost cartoonish visual language. The resemblance to the actual animal in the source photo tends to dissolve, leaving something that feels more like a generic AI illustration than an animated version of your specific pet. For reference, here’s the image I shared.
The custom prompt mode is where the feature really stood out. Using a before-and-after pair of photos of a succulent, one nearly bare and the other with fresh leaves beginning to sprout, I asked it to generate a video of the leaves growing, and it delivered. What it added beyond that was the more interesting part: the pot rotated gently as the camera slowly zoomed out, giving the whole clip a cinematic quality.
While my experience was largely positive, there are two things worth noting before you pull the trigger on the Honor 600 Pro to try this feature. The Generate button currently shows a “Limited-time trial” label, so fees will likely apply down the line, which might vary depending on your region. The feature is also limited to 10 generations per day, which is fine for casual use but may feel restrictive if you plan to use it heavily.
A few rough edges
The add audio feature worked fine in most cases, but things got inconsistent when human subjects were involved. In a couple of generated clips, subjects appeared to be speaking Chinese despite the device language being set to English. In some cases, East Asian facial features were applied to subjects in the photos.
These are likely artifacts of the model’s training data, which appears skewed toward East Asian users. However, this should improve over time as the model gets exposure to more diverse inputs. This isn’t a deal-breaker, but still worth knowing about if you plan to use the feature with photos of people.
Overall, Honor’s Image to Video 2.0 is a meaningful upgrade over what launched with the 400 series. The custom prompts, multi-image support, and template library push it from novelty into something with genuine creative utility. It’s not flawless, but when it works, it really works.
